arXiv cs.CL· Hang Su, Zequn Liu, Jianghan Chao, Qiuyu Tian, Xuesong Lu, Yingce Xia·· 12 小时前AI 评分34
PersonaFactor：面向 LLM 个性化的动态选择偏好因子框架
PersonaFactor: Preference Factors with Dynamic Selection for LLM Personalization
AI 导读
PersonaFactor 框架从异构用户交互中提炼出 13 个可复用、可解释的个性化因子，并动态选取与用户历史和当前请求相关的因子引导生成。在四个个性化基准和四个骨干模型上，Dynamic PersonaFactor 在所有报告指标上均取得最佳分数。表征分析显示，PersonaFactor 与用户历史诱导出部分重叠的低维表征，二者联合使用带来互补性偏移。
正文 · 原文
Abstract:Although users may express different personalization needs across tasks, these task-specific behaviors can reflect shared underlying preference mechanisms. Existing LLM personalization methods often focus on surface-level interactions or undifferentiated histories, overlooking the shared mechanisms underlying these behaviors. We introduce PersonaFactors, a framework that distills reusable and interpretable personalization factors from heterogeneous user interactions. We construct a library of thirteen PersonaFactors by extracting evidence-grounded rationales and clustering recurring patterns across tasks, and develop a dynamic factor-guided generation framework that selects factors relevant to each user's history and current request. Across four personalization benchmarks and four backbone models, Dynamic PersonaFactor achieves the best score on every reported metric. Representation analyses further show that PersonaFactors and user history induce partially overlapping low-dimensional representations, and that their joint use yields complementary shifts. Taken together, these findings suggest that factor-based and history-based personalization can mutually reinforce each other.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.16372 [cs.CL] |
| (or arXiv:2610.16372v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.16372 arXiv-issued DOI via DataCite |
Submission history
From: Hang Su [view email]
[v1]
Mon, 12 Oct 2026 08:23:23 UTC (3,770 KB)
[v2]
Tue, 20 Oct 2026 05:20:07 UTC (3,770 KB)
来源：arXiv cs.CL · arxiv.org